mirror of
https://github.com/storytold/pdfcraft.git
synced 2026-10-24 10:08:00 -06:00
PiohrunandClaude Opus 5.5
ba01244987
* M6: calculations read their notation only from a comment, and a regex after a condition as a regex Two findings from CodeRabbit's review of #3550, both reproduced first: - parse_calculate and calculation_sites found BVCALC and EVCALC anywhere in the script, so `var s = "BVCALC a EVCALC";` read as the notation `a`: renaming `a` rewrote the string, and the form calculated `a` instead of running the script. The lexer now keeps the text of each comment, and the notation is read from the first comment that holds BVCALC and the EVCALC after it (notation_range), by both the Calculate tab and a rename. A line comment counts, and a comment that never ends runs to the end of the script, as before. - A `/` after any `)` divided, so after the condition of `if`, `while`, `for` or `with` a regular expression was read as code: renaming `a` rewrote `if (flag) /getField("a")/.test(x)`, and in `if (flag) /"/.test(x)` the quote started a string that hid the script's real getField calls. The lexer now notes which `(` opened such a condition (not a method of that name, `o.if(x)`), and a `/` after its `)` starts a regular expression. af::tests::notation_is_read_only_from_a_comment, af::tests::a_regex_after_a_condition_is_not_read_as_code and tests::a_script_that_spells_the_markers_in_a_string_runs_as_a_script fail without the fix. Parity, ROADMAP.md and docs/roadmap.md updated. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z * M6: form.props-calculate notes that a regex right after a brace is still read as code From CodeRabbit's review of this PR. The lexer still divides after `}`, which ends a block as often as an object literal, so in `if (x) { } /"/.test(s); event.value = getField("a").value;` the quote starts a string and the getField is hidden from a rename and from the calculation order. Checked with a scratch test before writing it down. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z * M6: a line comment ends at any line terminator, so code after `\r` is still read From CodeRabbit's review of this PR. A `//` comment ran to the next `\n`, so with `\r` line endings it hid the rest of the script: a rename missed the getField calls after it (since #3550), and a string after it that spells BVCALC ... EVCALC read as the notation. A `/` taken for the start of a regular expression likewise looked past `\r` for its end, and could swallow the next line's code. Both now stop at any of JavaScript's line terminators (`\n`, `\r`, U+2028, U+2029; line_ends_at), and U+2028 and U+2029 between tokens are skipped rather than read as the start of a word. af::tests::a_line_comment_ends_at_any_line_terminator fails without each of the three changes: with `\r`, the string read as notation `a`; with the comment fixed but not the regex scan, `i++ / 2;\r` swallowed the getField after it; without skipping U+2028, the getField after it was not a call. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z * M6: a word, number or regex's flags end at U+2028 and U+2029 From CodeRabbit's review of this PR. The scans for an identifier, a number and a regular expression's flags took every byte past ASCII as part of a word, U+2028 and U+2029 included, so in `x\u{2028}getField("a")` the getField was part of one word and not a call: a rename missed it, and the calculation order didn't read it. `word` now takes a position and is false where line_ends_at is true, so the three scans share one rule. The arm that skips U+2028 and U+2029 between tokens stays: without it, the punctuation arm would step one byte into the character. af::tests::a_token_ends_at_a_line_separator fails without the fix, for a word, a number and a regex's flags alike; it and a_line_comment_ends_at_any_line_terminator fail without the skipping arm. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z * M6: the `while` that ends a `do` loop counts as a condition, said and tested A `/` after the `)` of a do-while starts the next statement, so it starts a regular expression there too. The comment on BEFORE_CONDITION says so, and a_regex_after_a_condition_is_not_read_as_code covers it (from review). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>